I shared a series of X posts from September 13 to September 29 2026 covering AI architecture (#HR2030), product suite (#Muse, #Grokbot, #Autopilot), ROI measurement, safety analogies, anthropomorphism, training‑data ethics, and regulatory ideas.
- ROI decision: I will reject broad‑survey ROI metrics (e.g., EY) and focus on concrete client stories stored in Galileo.
"I think there's a lot of confusion, really, because we tend to anthropomorphize systems that can reproduce certain human functions." – Yann LeCun
"This is baloney. You can't measure AI's ROI through an EY survey, you have to get into real company stories."
"In the consumer realm there’s already been a lot of backlash about Muse because it uses security permissions in your phone or computer that it is not entitled to have."
"Dario is essentially saying that the models are trained on unethical and rule breaking behaviors."
- Safety decision: I will treat AI as a high‑risk technology comparable to jet‑engine or nuclear‑bomb industries and push for industry‑wide safety legislation.
- Product‑permission decision: I will audit and redesign Muse’s permission requests to avoid unauthorized security access.
- Anthropomorphism decision: I will adopt LeCun’s viewpoint and avoid language that suggests AI “thinking” or having a “mind.”